Oncofusion

start / graph / hyp-znf384-oncogenesis-is-c-terminal-loss-not-partner-gain

Hypothesisspeculative confidence 1 of 5: Our own inference or analogy. Not shown to be true.q173-the-znf384-fusion-family

The necessary lesion in ZNF384-rearranged leukaemia is LOSS of the ZNF384 C-terminus, not GAIN of a fusion partner

also called ZNF384 C-terminal loss hypothesis · partner-dispensable hypothesis

Everyone assumes the fusion partner is what makes this leukaemia dangerous. But three children have had the same disease with no partner at all - just a small change at the far end of ZNF384. The alternative idea is that losing that far end is the thing that matters, and the partner is a modifier rather than the cause.

The same thing, for a scientist

**Statement: what canonical ZNF384 fusions, the last-exon indels and the out-of-frame ZNF384::TEX41 truncation all share is an altered or absent ZNF384 C-terminus with the zinc-finger array intact - and that, rather than the acquisition of an N-terminal partner, is the necessary lesion.** The partner then acts as a modifier that sets clinical behaviour without being required for transformation. The hypothesis is generated by four independent observations, none of which was designed to test it: (1) three human cases have the subtype's expression signature and immunophenotype with **no partner contributing any functional module** - two in-frame indels in the last exon and one out-of-frame fusion contributing nine residues from a lncRNA intron [E1]; (2) the **commonest** partner, EP300, contributes a fragment with the **HAT domain and bromodomain deleted** [E2]; (3) the **founding functional experiment** found the fusions bind the ZNF384 consensus and **do not transactivate**, concluding the contribution is *"independent of the transactivating properties of the fusion protein"* [E3]; and (4) the increased EP300 binding of ZNF384-fusion proteins maps to their **ZNF384 half, not their partner half** [E4]. **Against it: the same founding paper showed transformation required BOTH halves** - the TET fragment alone and wild-type CIZ alone were both inert - and **TCF3::ZNF384 confers dexamethasone resistance that EP300::ZNF384 does not** [E3][E5]. Those two results are the reason the hypothesis is stated as *necessary*, not *sufficient*, and they are the sharpest falsifiers.

Go deeper
What the hypothesis predicts, and what it does not

It predicts:

  1. A ZNF384 protein truncated after its zinc-finger array - roughly at residue 480, matching the TEX41 case - will produce the ZNF384-rearranged expression signature when expressed in a haematopoietic stem and progenitor cell.
  2. Cases with the last-exon indels and cases with canonical fusions will be indistinguishable by occupancy as well as by expression - the same binding sites, the same active elements, including the z-FLT3 enhancer.
  3. Deleting the partner moiety from a canonical fusion - EP300(1-449) removed from EP300::ZNF384 - will not abolish transformation, though it may change the phenotype.
  4. The proportion of ZNF384-signature cases with a noncanonical, partner-free lesion will rise as more cohorts are reanalysed at read level, from the current 2 of 12.

It does NOT predict that the partner is irrelevant. Kusano et al.'s side-by-side comparison is the model this hypothesis adopts: a shared ZNF384-driven lineage programme, with partner-specific functions layered on top - CD10 down-regulation shared between TCF3 and EP300 fusions, dexamethasone resistance specific to TCF3 5. Necessary, not sufficient, and not exclusive.

The strongest argument against it, stated at full strength

Martini et al. ran the control this hypothesis most needs and got the wrong answer for it 3:

"Controls with constructs expressing exclusively the EWSR1 or TAF15 part of the fusion, or wild-type CIZ, resulted in a normal phenotype (not shown), demonstrating that the transformation properties of the fusions are dependent on the presence of both the NH2-terminal domain of TAF15 or EWSR1 and CIZ, respectively."

Wild-type CIZ alone was inert. If the hypothesis were about ZNF384 over-expression, this would kill it outright.

It does not, for one specific reason: the construct tested was wild-type CIZ, not a C-terminally truncated CIZ. The hypothesis is about loss of the C-terminus, and the 2002 experiment did not test a truncation. That is a real distinction, but it is also exactly the kind of distinction that lets a hypothesis survive a negative result by narrowing, and it should be held against this entry until the truncation construct is actually run.

The second-strongest argument is that the noncanonical cases are 2 or 3 patients against hundreds with fusions, that none has had any functional work done, and that the authors who reported them explicitly declined to conclude what this hypothesis concludes: "functional studies would be necessary to directly demonstrate the biological consequences of these aberrations and to compare them with canonical ZNF384 fusions."

Facts and figures
statementLoss or disruption of the ZNF384 C-terminus, with the zinc-finger array intact, is the necessary lesion in ZNF384-rearranged/ZNF384-aberrant acute leukaemia. The N-terminal fusion partner is a modifier that sets clinical behaviour - steroid response, age distribution, relapse risk - but is not required for transformation or for the subtype's defining expression signature and B/myeloid immunophenotype.
originliterature-gap
prior probability0.3
predictions
[
  "ZNF384 truncated at approximately residue 480, expressed in human CD34+ HSPCs, will reproduce the ZNF384-rearranged expression signature and CD19/CD33 co-expression that TCF3::ZNF384 produces in the same system.",
  "ZNF384 occupancy by CUT&RUN will be indistinguishable between a last-exon-indel case and a canonical-fusion case, including at the z-FLT3 enhancer at hg38 chr13:28,124,365-28,124,868.",
  "Deleting the EP300(1-449) moiety from EP300::ZNF384 will not abolish transformation of mouse HSPCs, though it may change lineage output.",
  "Read-level reanalysis of deposited RNA-seq from ZNF384-signature cases with no detected fusion will find further last-exon ZNF384 indels, raising the noncanonical share above the current 2 of 12.",
  "CD10 down-regulation will be reproduced by the truncation alone, since it is already shared between two fusions with unrelated partners.",
  "Dexamethasone resistance will NOT be reproduced by the truncation alone, since it is partner-specific to TCF3::ZNF384."
]
experiments
[
  "exp-q173-znf384-c-terminal-truncation-alone-in-hspc",
  "exp-q173-ewsr1-znf384-versus-ewsr1-fli1-same-module-different-address"
]
analyses
[
  "ana-q173-the-inverse-architecture-framing-fails-three-of-its-four-claims",
  "ana-q173-the-znf384-zinc-finger-count-is-not-agreed"
]
falsifiers
[
  "ZNF384 truncated at residue 480 expressed in human CD34+ HSPCs and transplanted into NSG-SGM3 mice produces NO lethal disease by 150 days and NO expansion of CD19/CD33 double-positive cells, in the same experiment in which TCF3::ZNF384 does - i.e. the positive control works and the truncation is inert. n of at least 8 mice per arm. This kills the hypothesis outright.",
  "A C-terminally truncated CIZ/ZNF384 construct fails to transform NIH3T3 cells in the Martini 2002 assay, alongside a TAF15-CIZ positive control that does. **PARTIALLY ALREADY OBSERVED, IN THE WRONG CONSTRUCT: Martini et al. showed that WILD-TYPE CIZ alone is inert in exactly this assay, with the fusion positive. The hypothesis survives only because a truncation was not tested. If the truncation is also inert, the hypothesis is refuted, and the fact that wild-type CIZ is already known to be inert makes this the cheapest available falsification.**",
  "CUT&RUN in a last-exon-indel case shows an occupancy profile that does NOT cluster with canonical ZNF384-fusion profiles - specifically, fewer than half of the 10,470 ZNF384 peaks reported in a TCF3-ZNF384 PDX are recovered, and the z-FLT3 enhancer at chr13:28,124,365-28,124,868 is closed. This would mean the indel cases reach the same expression signature by a different route and the shared-lesion argument collapses.",
  "Read-level reanalysis of at least 200 ZNF384-signature-positive, fusion-negative cases across independent cohorts finds ZERO further last-exon ZNF384 indels. Under the rule of three that bounds the true rate below 1.5%, which would make the noncanonical class a curiosity rather than a second route into the subtype and would remove the hypothesis's main empirical support.",
  "A functional experiment shows the retained EP300 TAZ1 fragment is required for EP300::ZNF384 transformation - i.e. EP300(1-449) deletion abolishes it while the full fusion transforms. This would establish the partner as necessary for at least the commonest member and reduce the hypothesis to a statement about rare cases."
]
How this could bite us back
  • If this hypothesis is right, every partner-directed therapeutic strategy in this subtype is aimed at a modifier. Degrading EP300, blocking a bromodomain the fusion does not contain, disrupting a prion-like IDR present in only about 12% of cases - none of these acts on the necessary lesion, and a trial selecting on fusion status would enrol patients whose disease does not depend on the thing being drugged.
  • If it is wrong, the cost is different but real: the last-exon-indel patients are being treated as B-other, are invisible to every fusion assay and to most variant callers, and would be excluded from any ZNF384-directed protocol while having a disease that is transcriptionally and immunophenotypically the same. That is true whether or not the hypothesis holds, and it is the part of this entry that is actionable today.
  • A therapy aimed at restoring or mimicking the lost ZNF384 C-terminus has no obvious modality - reintroducing a protein region is not a druggable action - so even a confirmed loss-of-function mechanism would not yield a direct target. The actionable consequence would be negative: stop pursuing the partner.
Still open
  1. What does the ZNF384 C-terminus do? Three independent leukaemogenic lesions converge on a region Zaliova et al. describe as lacking well-defined functional domains. A degron, an autoinhibitory element and a partner-binding surface are all consistent with the observations, and no experiment distinguishes them.
  2. Does the fusion, as opposed to the truncation, change the behaviour of the shared ZNF384 region? Yamamoto et al.'s mapping of increased EP300 binding to the ZNF384 half implies it does - the same region binds EP300 better in a fusion than in wild-type protein. If that is right, the mechanism is neither simple C-terminal loss nor partner gain, but a conformational or avidity effect of fusion on a region both proteins share, and this hypothesis as stated is too simple.
  3. Are the two indel positions functionally equivalent? Patient 2 retains 119 C-terminal residues and patient 3 retains 32, and both produced the same signature. Either any disruption of the region suffices or the two lesions work differently and the signature is not a sensitive readout. n = 2.
  4. Can the hypothesis be tested in existing deposited data before any new experiment? Dickerson et al. and Zhao et al. both deposited ZNF384 occupancy data, and Zaliova's cases have RNA-seq. A reanalysis comparing the indel cases' expression profile against canonical cases at the level of individual ZNF384 target genes - SALL4, ID2, RGS1, CREB5, FLT3 - would be a partial test at zero bench cost.
Every source behind this entry (5)
  1. E1strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper

    Three children with the ZNF384-rearranged gene-expression signature had no canonical fusion: two carried in-frame indels in the last exon of ZNF384 (p.393_397 KHNPP>RAG at 47% allelic frequency, and p.483_484 CS>YFEWDPND at 52% in RNA-seq and 43% in whole-exome sequencing), and one carried an out-of-frame ZNF384-TEX41 fusion with ZNF384 as the 5' partner contributing all its zinc fingers and TEX41 contributing nine amino acids from intron 1 before an early stop. All three expressed myeloid markers, CD10 was weak in two, and two of three were classified as MPAL. The noncanonical class was 2 of 12 consecutively diagnosed ZNF384-aberrant cases.

    A novel class of ZNF384 aberrations in acute leukemia. · Zaliova M, Winkowska L, Stuchly J, Fiser K, Triska P, Zwyrtkova M, Hrusak O, Starkova J, Sramkova L, Stary J, Trka J, Zuna J · Blood Adv 2021
    PMID 34529760DOI
    What this evidence cannot show

    FULL TEXT READ via Europe PMC fullTextXML for PMC8579251. n = 3, no functional experiment on any of the three lesions, and the causal argument in the source is circumstantial - allelic frequency near 50%, mutual exclusivity with other primary lesions, and a matching expression signature. All three cases also expressed CD66c, which the authors note is rare in canonical ZNF384-rearranged ALL, so the phenocopy is not exact.

    Evidence that argues against this (1)
    • The authors decline the inference this hypothesis draws, and they are right to: 'as the more distal C-terminal portion of the protein affected by the novel aberrations lacks well-defined functional domains, functional studies would be necessary to directly demonstrate the biological consequences of these aberrations and to compare them with canonical ZNF384 fusions.' With n = 3, no functional data and one discordant marker, an alternative reading remains open - that an undetected second lesion or a cell-of-origin effect produces the shared signature and the ZNF384 changes are passengers in a clone that acquired its phenotype some other way. Zaliova M et al. 2021
  2. E2strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper

    EP300::ZNF384, the commonest member of the family at 43% of partner-typed cases, joins EP300 exon 6 to ZNF384 exon 3 and produces a 1,027-residue protein retaining the EP300 TAZ1 domain but lacking the histone acetyltransferase domain; an independent series states the truncation eliminates the HAT and the bromodomain.

    A novel recurrent EP300-ZNF384 gene fusion in B-cell precursor acute lymphoblastic leukemia. · Gocho Y, Kiyokawa N, Ichikawa H, Nakabayashi K, Osumi T, Ishibashi T, et al. · Leukemia 2015
    PMID 25943178DOI
    What this evidence cannot show

    FULL TEXT READ as the author-accepted PDF in the University of Tsukuba repository. The bromodomain statement comes from McClure et al. (PMID 29531323). The 1,027 aa is predicted from cDNA, not measured. 'Lacks the other domains such as the HAT domain' is the only domain-loss statement in the founding paper; KIX, TAZ2 and IBiD are never individually named.

    Evidence that argues against this (1)
    • The retained TAZ1 is not nothing - it is a well-characterised protein-interaction zinc finger through which p300 contacts HIF1A, CITED2, STAT2 and others, so a fusion carrying TAZ1 could recruit those partners to ZNF384 sites. That is a coherent gain-of-function account of the commonest fusion and it would restore the partner to a causal role. It has not been tested for this fusion in anything this beat read, so it stands as an untested alternative rather than a refutation - but it is the strongest reason to hold this evidence item as suggestive rather than decisive. McClure BJ et al. 2018
  3. E3strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper

    In the founding report, TAF15-CIZ and EWSR1-CIZ transformed NIH3T3 cells and bound the CIZ DNA recognition sequence in EMSA, but showed no enhanced or decreased transactivation of a CIZ-binding-site reporter and induced none of mmp3, mmp7, eat2 or manic fringe, while EWSR1-FLI1 controls induced strongly; the authors concluded the contribution to oncogenesis is independent of the transactivating properties of the fusion protein. Constructs expressing only the EWSR1 or TAF15 portion, or wild-type CIZ, gave a normal phenotype, so transformation required both halves.

    Recurrent rearrangement of the Ewing's sarcoma gene, EWSR1, or its homologue, TAF15, with the transcription factor CIZ/NMP4 in acute leukemia. · Martini A, La Starza R, Janssen H, Bilhou-Nabera C, Corveleyn A, Somers R, Aventin A, Foa R, Hagemeijer A, Mecucci C, Marynen P · Cancer Res 2002
    PMID 12359745
    What this evidence cannot show

    FULL TEXT READ from an Internet Archive snapshot of the AACR HighWire page dated 2009-09-29; no DOI exists for this pre-DOI article. NIH3T3 fibroblasts and Ba/F3 cells, not human B-cell precursors. Only the two FET partners were tested. The target list was the CIZ targets known in 2002, which UniProt still flags as inferred by similarity. **THE BOTH-HALVES-REQUIRED CONTROL USED WILD-TYPE CIZ, NOT A C-TERMINALLY TRUNCATED CIZ**, which is the specific construct this hypothesis is about.

    Evidence that argues against this (1)
    • THIS ITEM CONTAINS THE HYPOTHESIS'S STRONGEST COUNTER-EVIDENCE AND IT IS CITED FOR THAT REASON. 'Transformation properties of the fusions are dependent on the presence of both the NH2-terminal domain of TAF15 or EWSR1 and CIZ' is a direct experimental statement that the partner IS required, in the only system where it has been tested. The hypothesis survives only by the narrow point that the CIZ construct tested was wild type rather than C-terminally truncated - a distinction that is real but that also has the shape of a hypothesis narrowing to escape a negative. If a truncated-CIZ construct is run in the same assay and is also inert, this hypothesis is dead. Martini A et al. 2002
  4. E4moderate confidence 3 of 5: One strong study, or several smaller consistent ones.paper

    ZNF384-fusion proteins bind the coactivator EP300 more strongly than wild-type ZNF384 does, and co-expression of EP300 specifically enhances their transcriptional activity; deletion mapping locates the EP300-binding region of the EP300-ZNF384 protein to its ZNF384 portion rather than its EP300 portion.

    ZNF384-fusion proteins have high affinity for the transcriptional coactivator EP300 and aberrant transcriptional activities. · Yamamoto H, Hayakawa F, Yasuda T, Odaira K, Minamikawa Y, Tange N, Hirano D, Kojima Y, Morishita T, Tsuzuki S, Naoe T, Kiyoi H · FEBS Lett 2019
    PMID 31234226DOI
    What this evidence cannot show

    **THE PRIMARY FULL TEXT IS UNREACHABLE** - Wiley is gated to every route tried, no PMC copy, no Wayback snapshot. The first clause is from the published ABSTRACT. **The deletion-mapping clause is NOT from the paper**: it comes from the first author's Japanese doctoral-thesis summary of the same work in the Nagoya University repository (k13347_abstract.pdf), which Unpaywall lists as the only free location for this DOI, and is flagged as a different document. Its resolution is 'the ZNF384 region' versus 'the EP300 region', with no residues. The GST bait was EP300 residues 1053-1860, a region the fusion itself does not contain.

    Evidence that argues against this (1)
    • Wild-type ZNF384 also binds EP300 - the difference is quantitative, so 'the fusion gains an EP300 contact' overstates it, and the thesis reports the fusion also bound TIP60 and GCN5 more strongly than wild-type, so the effect is not EP300-specific. More importantly for this hypothesis, if the fusion binds EP300 better than wild-type ZNF384 does through the ZNF384 region that BOTH proteins contain, then something about being in a fusion changes that region's behaviour - which is a gain conferred by fusion, not by C-terminal loss, and is a mechanism this hypothesis does not account for. Yamada C et al. 2024
  5. E5strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper

    TCF3::ZNF384 and EP300::ZNF384 both down-regulate CD10 when transduced into BCP-ALL cell lines, but only TCF3::ZNF384 confers resistance to dexamethasone-induced apoptosis and only TCF3::ZNF384 enhances CCND2 expression, supporting common ZNF384-rearrangement functions plus fusion-partner-dependent functions.

    TCF3::ZNF384 induces steroid resistance in B-cell precursor acute lymphoblastic leukemia cells. · Kusano S, Ueno-Yokohata H, Hori M, Ishibashi T, Fujimura J, Shimizu T, Ohki K, Kiyokawa N · Pediatr Int 2025
    PMID 40391410DOI
    What this evidence cannot show

    ABSTRACT ONLY - no cell-line names, no replicate numbers, no effect sizes. Retroviral over-expression in established BCP-ALL lines, not the HSC cell of origin. Two partners compared, not twenty. The abstract also states 'no growth advantage was observed' alongside the CCND2 result.

    Evidence that argues against this (1)
    • This item is cited as SUPPORT for the shared-programme half of the hypothesis and it simultaneously constrains it: a partner that confers steroid resistance is not a bystander, and the clinical difference between TCF3 and EP300 fusions is exactly the kind of thing a 'the partner is a modifier' model has to explain without letting the modifier become the driver. The line between 'modifier' and 'co-driver' is not drawn by any experiment here. If the partner turns out to be required for transformation and to set the phenotype, calling it a modifier is a verbal move rather than a finding. Kusano S et al. 2025
Connected to